Axo: Tolerating Delay Faults in Real-Time Systems

We address delay faults: faults that cause a software component to take more time for completing an action than a given deadline. Such faults are particularly of interest in realtime mission-critical control applications that use general-purpose computing platforms to compute setpoints. A violation of realtime constraints associated with setpoints can result in failure. Existing benign and Byzantine fault-tolerance architectures do not tolerate delay faults. We discuss the challenges involved in tolerating such faults. Then, we list the requirements on the realtime systems that pave the way for our solution: Axo. We describe how Axo masks delay faults, and we conclude with open issues.